summaryrefslogtreecommitdiff
path: root/compiler
Commit message (Collapse)AuthorAgeFilesLines
* generalize GHC.Cmm.Dataflow to work over any node typewip/nr/polymorphic-dataflowNorman Ramsey2021-12-071-31/+46
| | | | | | See #20725. The commit includes source-code changes and a test case.
* More permissive parsing of higher-rank type IPssheaf2021-12-071-1/+1
| | | | | | | | | | | | | | | | | The parser now accepts implicit parameters with higher-rank types, such as `foo :: (?ip :: forall a. a -> a) => ...` Before this patch, we instead insisted on parentheses like so: `foo :: (?ip :: (forall a. a -> a)) => ...` The rest of the logic surrounding implicit parameters is unchanged; in particular, even with ImpredicativeTypes, this idiom is not likely to be very useful. Fixes #20654
* compiler: Eliminate accidental loop in GHC.SysTools.BaseDirBen Gamari2021-12-071-2/+2
| | | | | | | | As noted in #20757, `GHC.SysTools.BaseDir.findToolDir` previously contained an loop, which would be triggered in the case that the search failed. Closes #20757.
* Revert "Data.List specialization to []"Matthew Pickering2021-12-031-4/+1
| | | | | | | | | | This reverts commit bddecda1a4c96da21e3f5211743ce5e4c78793a2. This implements the first step in the plan formulated in #20025 to improve the communication and migration strategy for the proposed changes to Data.List. Requires changing the haddock submodule to update the test output.
* Require all dirty_MUT_VAR callers to do explicit stg_MUT_VAR_CLEAN_info ↵nineonine2021-12-022-5/+15
| | | | comparison (#20088)
* CmmToC: Cast possibly-signed results as unsignedBen Gamari2021-12-021-10/+40
| | | | | | | C11 rule 6.3.1.1 dictates that all small integers used in expressions be implicitly converted to `signed int`. However, Cmm semantics require that the width of the operands be preserved with zero-extension semantics. For this reason we must recast sub-word arithmetic results as unsigned.
* CmmToC: Always cast arguments as unsignedBen Gamari2021-12-021-4/+28
| | | | | | | As noted in Note [When in doubt, cast arguments as unsigned], we must ensure that arguments have the correct signedness since some operations (e.g. `%`) have different semantics depending upon signedness.
* CmmToC: Zero-extend sub-word size resultsBen Gamari2021-12-021-6/+41
| | | | | As noted in Note [Zero-extending sub-word signed results] we must explicitly zero-extend the results of sub-word-sized signed operations.
* CmmToC: Fix width of shift operationsBen Gamari2021-12-021-3/+15
| | | | | Under C's implicit widening rules, the result of an operation like (a >> b) where a::Word8 and b::Word will have type Word, yet we want Word.
* nativeGen/aarch64: Fix handling of subword valuesBen Gamari2021-12-023-83/+228
| | | | | | | | | | | | | | | | Here we rework the handling of sub-word operations in the AArch64 backend, fixing a number of bugs and inconsistencies. In short, we now impose the invariant that all subword values are represented in registers in zero-extended form. Signed arithmetic operations are then responsible for sign-extending as necessary. Possible future work: * Use `CMP`s extended register form to avoid burning an instruction in sign-extending the second operand. * Track sign-extension state of registers to elide redundant sign extensions in blocks with frequent sub-word signed arithmetic.
* cmm/opt: Fold away shifts larger than shiftee widthBen Gamari2021-12-021-2/+12
| | | | | This is necessary for lint-correctness since we no longer allow such shifts in Cmm.
* nativeGen/aarch64: Don't rely on register width to determine amodeBen Gamari2021-12-021-12/+16
| | | | | We might be loading, e.g., a 16- or 8-bit value, in which case the register width is not reflective of the loaded element size.
* cmm: Disallow shifts larger than shifteeBen Gamari2021-12-022-0/+19
| | | | | | | | | | | | Previously primops.txt.pp stipulated that the word-size shift primops were only defined for shift offsets in [0, word_size). However, there was no further guidance for the definition of Cmm's sub-word size shift MachOps. Here we fix this by explicitly disallowing (checked in many cases by CmmLint) shift operations where the shift offset is larger than the shiftee. This is consistent with LLVM's shift operations, avoiding the miscompilation noted in #20637.
* ncg/aarch64: Don't sign extend loadsBen Gamari2021-12-021-2/+2
| | | | | Previously we would emit the sign-extending LDS[HB] instructions for sub-word loads. However, this is wrong, as noted in #20638.
* cmm: narrow when folding signed quotientsBen Gamari2021-12-021-2/+2
| | | | | | | | | | | | | | | | | | Previously the constant-folding behavior for MO_S_Quot and MO_S_Rem failed to narrow its arguments, meaning that a program like: %zx64(%quot(%lobits8(0x00e1::bits16), 3::bits8)) would be miscompiled. Specifically, this program should reduce as %lobits8(0x00e1::bits16) == -31 %quot(%lobits8(0x00e1::bits16), 3::bits8) == -10 %zx64(%quot(%lobits8(0x00e1::bits16), 3::bits8)) == 246 However, with this bug the `%lobits8(0x00e1::bits16)` would instead be treated as `+31`, resulting in the incorrect result of `75`. (cherry picked from commit 94e197e3dbb9a48991eb90a03b51ea13d39ba4cc)
* nativeGen/x86: Don't encode large shift offsetsBen Gamari2021-12-021-1/+10
| | | | | | | | | | | Handle the case of a shift larger than the width of the shifted value. This is necessary since x86 applies a mask of 0x1f to the shift amount, meaning that, e.g., `shr 47, $eax` will actually shift by 47 & 0x1f == 15. See #20626. (cherry picked from commit 31370f1afe1e2f071b3569fb5ed4a115096127ca)
* Dump non-module specific info to file #20316Carrie Xu2021-12-013-16/+15
| | | | | | | | | - Change the dumpPrefix to FilePath, and default to non-module - Add dot to seperate dump-file-prefix and suffix - Modify user guide to introduce how dump files are named - This commit does not affect Ghci dump file naming. See also #17500
* rts: Refactor SRT representation selectionBen Gamari2021-12-012-9/+31
| | | | | | The goal here is to make the SRT selection logic a bit clearer and allow configurations which we currently don't support (e.g. using a full word in the info table even when TNTC is used).
* Use Monoid in hptSomeThingsBelowUsSylvain Henry2021-11-291-18/+15
| | | | | | | | | | It seems to have a moderate but good impact on perf tests in CI. In particular: MultiLayerModules(normal) ghc/alloc 3125771138.7 3065532240.0 -1.9% So it's likely that huge projects will benefit from this.
* TTG: replace Void/NoExtCon with DataConCantHappenKrzysztof Gogolewski2021-11-2913-147/+134
| | | | | | | | There were two ways to indicate that a TTG constructor is unused in a phase: `NoExtCon` and `Void`. This unifies the code, and uses the name 'DataConCantHappen', following the discussion at MR 7041. Updates haddock submodule
* Make ambient MinGW support a proper settingsJohn Ericson2021-11-274-26/+39
| | | | | | | Get rid of `USE_INPLACE_MINGW_TOOLCHAIN` and use a settings file entry instead. The CPP setting was originally introduced in f065b6b012.
* Allow keywords which can be used as variables to be used with ↵Matthew Pickering2021-11-261-1/+1
| | | | | | | | | | | | | OverloadedDotSyntax There are quite a few keywords which are allowed to be used as variables. Such as "as", "dependency" etc. These weren't accepted by OverloadedDotSyntax. The fix is pretty simple, use the varid production rather than raw VARID. Fixes #20723
* Reflect type change in the haddock commentKai Prott2021-11-261-1/+1
|
* Fix plugin type to GHC.Plugins.PluginKai Prott2021-11-261-1/+1
|
* Simplify printQualificationKai Prott2021-11-261-7/+2
|
* Remove duplicate importKai Prott2021-11-261-1/+0
|
* Improve error message for mis-typed plugins #20671Kai Prott2021-11-262-15/+25
| | | | Previously, when a plugin could not be loaded because it was incorrectly typed, the error message only printed the expected but not the actual type. This commit augments the error message such that both types are printed and the corresponding module is printed as well.
* compiler: Mark GHC.Prelude as Haddock no-homeBen Gamari2021-11-251-0/+1
| | | | This significantly improves Haddock documentation generated by nix.
* Allow boring class declarations in hs-boot filessheaf2021-11-251-1/+9
| | | | | | | | | | | | | | | | | | | | | | | There are two different ways of declaring a class in an hs-boot file: - a full declaration, where everything is written as it is in the .hs file, - an abstract declaration, where class methods and superclasses are left out. However, a declaration with no methods and a trivial superclass, such as: class () => C a was erroneously considered to be an abstract declaration, because the superclass is trivial. This is remedied by a one line fix in GHC.Tc.TyCl.tcClassDecl1. This patch also further clarifies the documentation around class declarations in hs-boot files. Fixes #20661, #20588.
* Add `llvmOptLevel` to `DynFlags` (#20500)Gergo ERDI2021-11-252-5/+7
|
* Add specific optimization flag for Cmm control flow analysis (#20500)Gergo ERDI2021-11-253-1/+4
|
* Add specific optimization flag for fast PAP calls (#6084, #20500)Gergo ERDI2021-11-253-1/+4
|
* Use `simplify` in non-optimizing build pipeline (#20500)Gergo ERDI2021-11-251-4/+2
|
* GHCi Debugger - Improve RTTIRoland Senn2021-11-251-0/+3
| | | | | | | When processing the heap, use also `APClosures` to create additional type constraints. This adds more equations and therefore improves the unification process to infer the correct type of values at breakpoints. (Fix the `incr` part of #19559)
* drop instance Semigroup InstalledModuleEnvPepe Iborra2021-11-251-1/+4
| | | | Instead, introduce plusInstalledModuleEnv
* Drop instance Semigroup ModuleEnvPepe Iborra2021-11-251-1/+0
| | | | There is more than one possible Semigroup and it is not needed since plusModuleEnv can be used directly
* Monoid instance for InstalledModuleEnvPepe Iborra2021-11-251-0/+3
|
* Don't use implicit lifting when deriving LiftMatthew Pickering2021-11-251-8/+20
| | | | | | | | | | | | | | | | It isn't much more complicated to be more precise when deriving Lift so we now generate ``` data Foo = Foo Int Bool instance Lift Foo where lift (Foo a b) = [| Foo $(lift a) $(lift b) |] liftTyped (Foo a b) = [|| Foo $$(lift a) $$(lift b) |] ``` This fixes #20688 which complained about using implicit lifting in the derived code.
* Misc cleanupKrzysztof Gogolewski2021-11-2512-58/+23
| | | | | | | | | | | | | | | | | | | | | * Remove `getTag_RDR` (unused), `tidyKind` and `tidyOpenKind` (already available as `tidyType` and `tidyOpenType`) * Remove Note [Explicit Case Statement for Specificity]. Since 0a709dd9876e40 we require GHC 8.10 for bootstrapping. * Change the warning to `cmpAltCon` to a panic. This shouldn't happen. If it ever does, the code was wrong anyway: it shouldn't always return `LT`, but rather `LT` in one case and `GT` in the other case. * Rename `verifyLinearConstructors` to `verifyLinearFields` * Fix `Note [Local record selectors]` which was not referenced * Remove vestiges of `type +v` * Minor fixes to StaticPointers documentation, part of #15603
* Enable UnboxedTuples in `genInst`, Fixes #20524CarrieMY2021-11-251-0/+3
|
* Compare types of recursive let-bindings in alpha-equivalenceChristiaan Baaij2021-11-256-146/+254
| | | | | | | | | | | | | This commit fixes #20641 by checking the types of recursive let-bindings when performing alpha-equality. The `Eq (DeBruijn CoreExpr)` instance now also compares `BreakPoint`s similarly to `GHC.Core.Utils.eqTickish`, taking bound variables into account. In addition, the `Eq (DeBruijn Type)` instance now correctly compares the kinds of the types when one of them contains a Cast: the instance is modeled after `nonDetCmpTypeX`.
* Ensure new Ct/evidence invariantChristiaan Baaij2021-11-254-31/+80
| | | | | | | | | | The `ctev_pred` field of a `CtEvidence` is a just a cache for the type of the evidence. More precisely: * For Givens, `ctev_pred` = `varType ctev_evar` * For Wanteds, `ctev_pred` = `evDestType ctev_dest` This new invariant is needed because evidence can become part of a type, via `Castty ty kco`.
* Revert "Convert lookupIdSubst panic back to a warning (#20200)"Matthew Pickering2021-11-251-6/+4
| | | | This reverts commit df1d808f26544cbb77d85773d672137c65fd3cc7.
* Correct retypechecking in --make modeMatthew Pickering2021-11-256-128/+290
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Note [Hydrating Modules] ~~~~~~~~~~~~~~~~~~~~~~~~ What is hydrating a module? * There are two versions of a module, the ModIface is the on-disk version and the ModDetails is a fleshed-out in-memory version. * We can **hydrate** a ModIface in order to obtain a ModDetails. Hydration happens in three different places * When an interface file is initially loaded from disk, it has to be hydrated. * When a module is finished compiling, we hydrate the ModIface in order to generate the version of ModDetails which exists in memory (see Note) * When dealing with boot files and module loops (see Note [Rehydrating Modules]) Note [Rehydrating Modules] ~~~~~~~~~~~~~~~~~~~~~~~~~~~ If a module has a boot file then it is critical to rehydrate the modules on the path between the two. Suppose we have ("R" for "recursive"): ``` R.hs-boot: module R where data T g :: T -> T A.hs: module A( f, T, g ) where import {-# SOURCE #-} R data S = MkS T f :: T -> S = ...g... R.hs: module R where data T = T1 | T2 S g = ...f... ``` After compiling A.hs we'll have a TypeEnv in which the Id for `f` has a type type uses the AbstractTyCon T; and a TyCon for `S` that also mentions that same AbstractTyCon. (Abstract because it came from R.hs-boot; we know nothing about it.) When compiling R.hs, we build a TyCon for `T`. But that TyCon mentions `S`, and it currently has an AbstractTyCon for `T` inside it. But we want to build a fully cyclic structure, in which `S` refers to `T` and `T` refers to `S`. Solution: **rehydration**. *Before compiling `R.hs`*, rehydrate all the ModIfaces below it that depend on R.hs-boot. To rehydrate a ModIface, call `typecheckIface` to convert it to a ModDetails. It's just a de-serialisation step, no type inference, just lookups. Now `S` will be bound to a thunk that, when forced, will "see" the final binding for `T`; see [Tying the knot](https://gitlab.haskell.org/ghc/ghc/-/wikis/commentary/compiler/tying-the-knot). But note that this must be done *before* compiling R.hs. When compiling R.hs, the knot-tying stuff above will ensure that `f`'s unfolding mentions the `LocalId` for `g`. But when we finish R, we carefully ensure that all those `LocalIds` are turned into completed `GlobalIds`, replete with unfoldings etc. Alas, that will not apply to the occurrences of `g` in `f`'s unfolding. And if we leave matters like that, they will stay that way, and *all* subsequent modules that import A will see a crippled unfolding for `f`. Solution: rehydrate both R and A's ModIface together, right after completing R.hs. We only need rehydrate modules that are * Below R.hs * Above R.hs-boot There might be many unrelated modules (in the home package) that don't need to be rehydrated. This dark corner is the subject of #14092. Suppose we add to our example ``` X.hs module X where import A data XT = MkX T fx = ...g... ``` If in `--make` we compile R.hs-boot, then A.hs, then X.hs, we'll get a `ModDetails` for `X` that has an AbstractTyCon for `T` in the the argument type of `MkX`. So: * Either we should delay compiling X until after R has beeen compiled. * Or we should rehydrate X after compiling R -- because it transitively depends on R.hs-boot. Ticket #20200 has exposed some issues to do with the knot-tying logic in GHC.Make, in `--make` mode. this particular issue starts [here](https://gitlab.haskell.org/ghc/ghc/-/issues/20200#note_385758). The wiki page [Tying the knot](https://gitlab.haskell.org/ghc/ghc/-/wikis/commentary/compiler/tying-the-knot) is helpful. Also closely related are * #14092 * #14103 Fixes tickets #20200 #20561
* Combine STG free variable traversals (#17978)nineonine2021-11-239-285/+255
| | | | | | | | | | | | | | | | | | | | | Previously we would traverse the STG AST twice looking for free variables. * Once in `annTopBindingsDeps` which considers top level and imported ids free. Its output is used to put bindings in dependency order. The pass happens in STG pipeline. * Once in `annTopBindingsFreeVars` which only considers non-top level ids free. Its output is used by the code generator to compute offsets into closures. This happens in Cmm (CodeGen) pipeline. Now these two traversal operations are merged into one - `FVs.depSortWithAnnotStgPgm`. The pass happens right at the end of STG pipeline. Some type signatures had to be updated due to slight shifts of StgPass boundaries (for example, top-level CodeGen handler now directly works with CodeGen flavoured Stg AST instead of Vanilla). Due to changed order of bindings, a few debugger type reconstruction bugs have resurfaced again (see tests break018, break021) - work item #18004 tracks this investigation. authors: simonpj, nineonine
* Add a warning for GADT match + NoMonoLocalBinds (#20485)Krzysztof Gogolewski2021-11-235-12/+39
| | | | | | | | | Previously, it was an error to pattern match on a GADT without GADTs or TypeFamilies. This is now allowed. Instead, we check the flag MonoLocalBinds; if it is not enabled, we issue a warning, controlled by -Wgadt-mono-local-binds. Also fixes #20485: pattern synonyms are now checked too.
* CmmSink: Be more aggressive in removing no-op assignments.Andreas Klebinger2021-11-231-3/+57
| | | | | | | | | | No-op assignments like R1 = R1 are not only wasteful. They can also inhibit other optimizations like inlining assignments that read from R1. We now check for assignments being a no-op before and after we simplify the RHS in Cmm sink which should eliminate most of these no-ops.
* Better wrapper activation calculationSimon Peyton Jones2021-11-222-34/+80
| | | | | | | | | | | | | As #20709 showed, GHC could prioritise a wrapper over a SPEC rule, which is potentially very bad. This patch fixes that problem. The fix is is described in Note [Wrapper activation], especially item 4, 4a, and Conclusion. For now, it has a temporary hack (replicating what was there before to make sure that wrappers inline no earlier than phase 2. But it should be temporary; see #19001.
* Include "not more specific" info in overlap msgsheaf2021-11-202-46/+152
| | | | | | | | | When instances overlap, we now include additional information about why we weren't able to select an instance: perhaps one instance overlapped another but was not strictly more specific, so we aren't able to directly choose it. Fixes #20542
* Use 'NonEmpty' for the fields in an 'HsProjection' (#20389)Zubin Duggal2021-11-2010-17/+57
| | | | | | | | | | | | T12545 is very inconsistently affected by this change for some reason. There is a decrease in allocations on most configurations, but an increase on validate-x86_64-linux-deb9-unreg-hadrian. Accepting it as it seems unrelated to this patch. Metric Decrease: T12545 Metric Increase: T12545
View Lorry Controller Status